Research On Reaction Modelling And Numerical Simulation About CVD System In Polysilicon Reduction Furnace

Since the worldwide energy situation becomes increasingly tense,much attention has been paid to solar energy,a green and renewable energy source.Being core process of solar PV industry,polysilicon is now into an extremely critical position in history.The modified Siemens process,those using tri-chlorinated silane as precursor,has dominated the global polysilicon production in terms of market share,leaving fluidized bed method far behind.This might be attributed to the ability of rapid capacity expansion while keeping its purity advantage.Because of the overwhelming market leadership of the Siemens process,research on Siemens process appears to be highly attractive in aspect of commercial and industrial potential.CFD has been introduced to the simulation of polysilicon reduction furnaces for a long time.Due to poor understanding of transfer and reaction process in reduction furnaces,many researchers have to make lots of simplifications to internal process in simulation.In this paper,a detailed reaction model is constructed first.The constructed reaction model is then validated by simulations of two-dimensional laboratory-scale reactors.Finally,the reaction model is combined with a three-dimensional CFD model to investigate transfer and reaction patterns in industrial-level reactors.Research findings are listed as follows:(1)Firstly,A new reaction model is constructed.In terms of thermodynamics,more accurate and reliable thermodynamic properties of gas phase species are obtained by quantum calculations.In terms of reaction kinetics,the architecture of surface reactions was reorganized.Data from experiment measurement is preferred when referring to desorption of key components.Results show that the conventional reaction model would show a significant and unreasonable performance degradation when inlet TCS molar fraction exceeds 20%,which is corrected in our new reaction model.The new reaction model also predicts a more reasonable response of deposition rate to operating pressure and is able to achieve reasonable deposition predictions under industrial-level operating conditions,which provided the theoretical basis for our subsequent CFD calculations.(2)Based on computational fluid dynamics,models of four different laboratory-scale reactors are constructed.These results suggest that the difference between new mechanism,conventional reaction mechanism and experimental measurements is quite small,and the largest deviation does not exceed 20%,with respect to film growth in ideal chamber.Considering Habuka deposition experiments,the new mechanism is in full agreement with the conventional mechanism model when inlet TCS mole fraction is less than 3%.However,the higher inlet TCS mole fraction,the larger deviation between the conventional mechanism and the experiment results.At 5% inlet TCS molar fraction,the deviation of new mechanism from the experimental values can be controlled under 1%,while that of conventional mechanism is 10.1%.From the view of Angermeier’s experiments,our new mechanism predicts the temperature window of overall deposition reaction in the range of1100 K～1325 K,which is consistent with the experiment,while temperature range obtained by the conventional mechanism is 1275 K～1400 K,with significant differences between the upper and lower limits.Meanwhile,the new mechanism also successfully captures the trend of decreasing deposition rate that occurs above 1425 K.Based on Habuka’s etch experiments,the conventional mechanism shows an extremely low etch rate close to 0.The deviation of the predicted etch rate from the experimental data by the new mechanism model is under 5%,and the overall reaction of etch about hydrogen chloride is first order.Silicon tetrachloride is the dominant component in tail gas which is in accordance with thermodynamic analysis results in the literature and mass spectra data of the tail gas given by Habuka.(3)A three-dimensional single silicon rod CFD model is constructed for industrial-level operating conditions.The results show that the top space of the reduction furnace is controlled by buoyancy while the bottom is dominated by convection effect.The deposition pattern varies between two regions.When surface temperature is low,the deposition is controlled by desorption of chlorine;at higher temperatures,it is mainly controlled by adsorption of chlorosilane.The transition between different control regime will be completed in the range of 1323 K～1373 K.The deposition rate in the region controlled by convection effect of inlet gas is lower when surface temperature falls behind1323 K and vice versa for surface temperature exceeds 1373 K.An excessively high Si-H ratio is detrimental to deposition and will result in a lower conversion rate.Considering deposition rate as well as uniformity,the best operating condition under 6 atm is at surface temperature 1373 K;inlet TCS gas flow rate around 550 kg/h(equivalent inlet gas velocity:60～70 m/s);and inlet TCS molar fraction around 30%.The study of the reduction furnace process in this paper provides a feasible technical routine for future simulation and optimization of the Siemens reactor.The proposed reaction model provides a theoretical basis for an in-depth understanding of the internal characteristics of the reduction process and gives a feasible future research route at the end,which has a respectable industrial value.